- Email: [email protected]
Modeling behavior control of golden apple snails at different temperatures
G Model ECOMOD 7353 No. of Pages 9
Ecological Modelling xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel
Modeling behavior control of golden apple snails at different temperatures Mi-Jung Bae a , Tae-Soo Chon b , Young-Seuk Park a,c, * a
Department of Biology, Kyung Hee University, Dongdaemun, Seoul 130-701, Republic of Korea Department of Biological Sciences, Pusan National University, Geumjeong, Busan 609-735, Republic of Korea c Department of Life and Nanopharmaceutical Sciences, Kyung Hee University, Seoul 130-701, Republic of Korea b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 1 April 2014 Received in revised form 15 October 2014 Accepted 15 October 2014 Available online xxx
The golden apple snail (Pomacea canaliculata) is a detrimental invasive pest of rice in Asia. Temperature is a critical factor to determine their invasiveness including the behavior and distribution of snails. In this study, the behavioral responses of golden apple snails acclimated to different water temperatures (15 C, 20 C, 25 C, and 30 C) were examined based on Markov chain analysis, Shannon entropy, and Random forest modeling. Markov chains revealed that most snails maintained their previous behavior at low temperatures, while behavior transition tended to be higher at high temperature. Shannon entropy was also dependent on temperature (low at low temperature and high at high temperature), indicating that snails maintained their previous behavior for a long time at low temperature regardless having motion and motionless behavior, whereas they changed continuously their behavioral types at high temperature. The Random forest model showed that Shannon entropy at low temperature was influenced by crawling at the bottom or at side of the aquarium (motion behavior), while clinging to the side of the aquarium or clinging to the side of the aquarium with stretching out antennae were important behaviors determining Shannon entropy at high temperature. Our results showed that snails controlled their behavior to reduce their thermal stress and maintain a stable internal state in harsh environment. With this mechanism, they are able to overwinter in the open fields with low temperature in Korea, resulting in the increase of potential damage in agricultural ecosystems. Therefore, further study on the development of adequate management system is required to avoid their invasiveness to open water systems and ecological impacts. Finally, Shannon entropy, Markov chain, and Random forest model are useful computation methods to quantify the effects of temperature changes on the behavior of golden apple snails. ã 2014 Elsevier B.V. All rights reserved.
Keywords: Pomacea canaliculata Behavior control Random forest model Markov chain Water temperature Shannon entropy
1. Introduction Invasion by exotic species is one of the main causes of the global extinction crisis, causing biodiversity loss and biotic homogenization (Mooney and Hobbs, 2000). Once successfully introduced and established in a new area, invasive species are difficult to be eradicated. Therefore, it is essential to understand the mechanisms and processes of biological invasions. The golden apple snail (Pomacea canaliculata) is an exotic species in Asian countries, introduced from South America, which has become a serious agricultural pest, especially to young rice (Halwart, 1994; Wada,
* Corresponding author at: Department of Biology, Kyung Hee University, Dongdaemun, Seoul 130-701, Republic of Korea. Tel.: +82 2 961 0946. E-mail addresses: [email protected] (M.-J. Bae), [email protected] (T.-S. Chon), [email protected] (Y.-S. Park).
2004; Cowie et al., 2006; Hayes et al., 2008). The golden apple snail was considered a pest in the Philippines in 1986 (Rejesus et al., 1990), and since then there was an annual economic loss of about 1200 million US dollars due to aquatic crop damage (Naylor, 1996). In Japan, the distribution range of this species has increased and, in 1999, ca. 13,000 ha of paddy fields have been damaged (Wada, 2004). In addition to the severe effects of rice paddy, the snail can cause major environmental changes, such as the depletion of macrophytes in natural wetlands (Carlsson et al., 2004). Hence, the golden apple snail is indicated as one of the 100 worst invasive species worldwide by the International Union for the Conservation of Nature (Lowe et al., 2000; Bang and Cho, 2008). Temperature is a critical factor for successful establishment and colonization of invasive species, including the golden apple snail, in new areas (Baker, 1998; EFSA, 2012), owing to its effects on survival (Seuffert and Martín, 2013), growth rate (Estebenet and Cazzaniga, 1992), reproduction (Seuffert et al., 2010), lung
http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020 0304-3800/ ã 2014 Elsevier B.V. All rights reserved.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
2
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
ventilation frequency (Seuffert and Martín, 2010), feeding rate (Seuffert et al., 2010), and behavior (Heiler et al., 2008; Seuffert et al., 2010; Seuffert and Martín, 2010). Snails that overwinter in paddy fields are more tolerant to cold conditions than those that do not overwinter (Wada and Matsukura, 2007). They usually remain inactive or bury in muddy bottoms during winter or periods of cold weather conditions (Damborenea, 1996; Seuffert et al., 2010; RDA, 2007). In areas with constant warm temperatures (e.g., 25 C), females of golden apple snails reproduce only once in their lifetime (i.e., semelparous), whereas in areas with fluctuating temperature conditions (e.g., 7–24 C), they have multiple reproductive periods (i.e., iteroparous) (Estebenet and Cazzaniga, 1992; Estebenet and Martín, 2002; Seuffert et al., 2010). High movement activity and crawling velocity are observed in areas with high water temperatures (Heiler et al., 2008). However, water temperatures above 30 C were seen to cause a decrease in activity and an increase in feeding time in laboratory experiments (Seuffert et al., 2010). Animals’ behavior and physiological responses are influenced by environmental factors. Therefore, understanding the behavior of invasive species is the first step in developing an effective control and management method. However, it is difficult to analyze behavioral data of organisms as they exhibit nonlinear and complex behaviors (Bae and Park, 2014). Therefore, the analysis of such data requires the use of appropriate analytical and modeling methods, including statistical and computational approaches (Chon et al., 2004). The Shannon information theory can be used to quantify variation in behavior, duration of behavior, and behavior complexity of organisms, using different sequences of behavioral types (Fig. 1). In fact, the Shannon entropy (Shannon and Weaver, 1949) has been widely applied in animal behavior analysis, e.g., in hermit crabs (Hazlett and Bossert, 1965), dolphins (McCowan et al., 1999), the rufous-bellied thrush (Da Silva et al., 2000), and the Japanese medaka (Fukuda et al., 2010). However, only a few studies focus on behavioral sequences (Gherardi and Pieraccini, 2004). A behavioral sequence can be modeled through the Markov chain, a discrete random process in which the probability of a system in the next time step depends only on the current state of the system and is independent of the previous
step (Guan et al., 2008). The Markov chain has been used to analyze landscape changes (e.g., Baker, 1989), population ecology (e.g., Seneta, 1966; Fujiwara and Caswell, 2002), and behavioral ecology (Lusseau, 2003). Since it quantifies the dependency between the present state and the state in the next time step, it can be used to compare the probability of behavioral transitions in response to different environmental conditions (Lusseau, 2003). Most studies dealing with behavior have either used short time observations (e.g., for 2 h) or only studied a certain type of behavior at a time, such as ventilation frequency or activity. To our knowledge, there are little quantitative researches on short time interval responses in behavior. Therefore, in this study, the behavioral response of golden apple snails to different water temperatures was examined based on short time interval data (i.e., every minute). Specifically, three steps of analyses were conducted: (1) the degree of behavior complexity at different water temperatures was compared based on Shannon entropy; (2) the duration of behavior as well as behavior transitions were compared through Markov chain analysis; and (3) the factors determining behavior complexity were evaluated based on Random forest model. 2. Materials and methods 2.1. Test organisms and behavioral observations Test organisms were obtained from a farm of golden apple snails (P. canaliculata) in Korea (http://www.gpwoolunge.co.kr/). A stock population was maintained in an aquarium with dechlorinated tap water (water temperature: 25 C 1 C; L16: D8) and a 3-cm layer of sand at the bottom, for at least 2 months before starting the experiments. In order to rule out irregular behavior (e.g., spawning and mating), only males were used. These were identified based on the presence of testis, observed through the translucent shell (Takeda, 1999), or by a humped operculum (Estebenet et al., 2006) and a body size (>25 mm) (Seuffert et al., 2010). The behavior of test organisms was observed in aquaria (30 cm 30 cm) filled with water to a depth of 15 cm and with a
Fig. 1. Flow chart of behavior data analyses of golden apple snails.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
3 cm sediment layer for two days. Test organisms have been acclimated to four different temperatures (15 C, 20 C, 25 C, and 30 C) for one month prior to the experiment. The amount of lettuce fed to the snails during experiment was estimated based on the differences of lettuce leaf area before and after the experiments. Lettuce was replaced once a day in aquaria set at 15 C, 20 C, and 25 C and twice a day in the aquarium set at 30 C due to higher consumption. In each experimental condition, the behaviors of 16 test organisms were observed. The following 12 behaviors were recorded at 1-min intervals by using a web cam (C905, Logitech) (Table 1): clinging on to the side of the aquarium (clinging); not moving with a closed operculum to the bottom of the aquarium (not-moving); floating with a closed operculum (floating); clinging to the side of the aquarium with stretching out antennae (clinging-NM); crawling at the bottom of aquarium (bottom-crawling); crawling at the side of aquarium (side-crawling); digging in the sand (digging); folding its foot to make a shape like a funnel (folding); falling from the side of aquarium (falling); swimming; breathing by lung ventilation (ventilating); and feeding on lettuce (feeding). They were categorized into three groups (i.e., motionless, semi-motionless and motion) based on the degree of movement and motion behaviors (Table 1). If organisms did not move, they were considered motionless, and if they did not display any motion except the movement of their antennae, they were considered semi-motionless; other behaviors were considered as motion. 2.2. Modeling procedure 2.2.1. Descriptive measures The time spent in each type of behavior was compared between the different water temperature treatments using the Kruskal– Wallis (K–W) test and the Dunn’s multiple comparison tests, with the software Statistica (ver. 7.0; StatSoft, 2004). The duration of a certain behavior was determined based on the amount of time that golden apple snails showed uninterruptedly that behavior. To evaluate the relationship between the duration of each behavior (log transformed) and its occurrence frequency (log transformed), we tested various curve fitting regressions using a software CurveExpert 1.3 (http://curveexpert.webhop.biz/), and chose Morgan-Mercer-Flodin (MMF, Morgan et al.,1975) which showed a high fitting performance. Among the 12 categories of behavior, the six dominant behaviors (i.e., clinging, clinging-NM, bottom-crawling, side-crawling, feeding, and not-moving) were considered in this analysis. 2.2.2. Markov chain The Markov chain (also called discrete-time Markov chain or Markov model) is a stochastic process in which the probability
3
associated with transitions among different behavioral states depends solely on the current state (Clark, 2007). A change in behavioral state in the system is called a “transition,” and the probabilities associated with various state changes are called “transition probabilities.” In the Markov chain, aij is the transition probability in the system from state i at time t to state j at time t + 1. If the system has a finite number of states (1, 2, . . . , n), the Markov chain can be characterized by a transition probability matrix denoted by A in Eq. (1) (Winston, 1994):
In the present study, the transition probability matrices of the 12 categories of behavior in the four water temperature treatments were compared. In order to make sure that our observation time was enough to generalize the behavioral transition probability, Markov chains of increasing observation time were computed (i.e., Markov chain computation based on observations of 1 h, 2 h, . . . , 48 h). Markov chains were calculated in MATLAB 7. The changes in transition probabilities according to the increasing observation time were shown as the form of a matrix plot, and the Markov chain based on a 2-day observation was selected to explain the transition probabilities. 2.2.3. Shannon entropy Shannon entropy (H) takes into account the occurrence probability of each behavior as shown in Eq. (2) (Gherardi and Pieraccini, 2004): H¼
n X pðiÞlog2 pðiÞ
(2)
i¼1
where p(i) is the occurrence probability of behavior i (i.e., the occurrence frequency) and N is the number of total behaviors (N = 12). In this study, Shannon entropy was calculated using a 10-min moving average of the behavior data registered every minute. The ‘entropy’ package in R (http://cran.r-project.org) was used for this analysis. 2.2.4. Random forest Random forest (RF) model is a machine learning technique based on a combination of a large set of decision trees (Breiman, 2001). RF has no assumption concerning the relationships between predictor and response variables (e.g., linear or nonlinear), and can be applied on continuous and categorical predictors (Prasad et al., 2006; Cutler et al., 2007; Hawkins et al., 2009). A randomly selected subset of predictor variables is applied to a randomly
Table 1 Behavior categories of Pomacea canaliculata. Category
Behavior
Acronym
Motionless
Clinging to the side of the aquarium Not moving with a closed operculum on the bottom of the aquarium Floating with a closed operculum
Clinging Not-moving Floating
Semi-motionless
Clinging to the side of the aquarium with stretching out antennae
Clinging-NM
Motion
Crawling at the bottom of the aquarium Crawling at the side of the aquarium Digging in the sand Folding its foot to make a shape like a funnel Falling from the side of the aquarium Swimming Breathing by lung ventilation Feeding on lettuce
Bottom-crawling Side-crawling Digging Folding Falling Ventilating Feeding
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
4
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
selected subset of samples, minimizing overfitting and producing unbiased estimates of modeling errors. To understand the effects of the previous behavior on the variations of behavior diversity (i.e., Shannon entropy), 12 behaviors at the current time step (t) were used in this study as the independent variables, and Shannon entropy at the following time steps (e.g., t + 1, t + 5 and t + 10) was used as the dependent variable. The ‘randomForest’ package (Liaw and Wiener, 2002) in R (http://cran.r-project.org) was used for this analysis, using the default setting of three training parameters (i.e., ntree, mtry, and node size) in the RF algorithm. The variable importance was calculated using the mean decrease in accuracy. For an easier interpretation, values of variable importance were rescaled as the proportion of each behavior; therefore, the sum of variable importance values from all categories was 1.
other temperatures. Clinging-NM (22.0%) and motion behaviors including crawling (side-crawling: 33.5% and bottom-crawling: 18.3%), folding (1.3%), falling (2.2%), swimming (0.1%), and ventilating (3.3%) were the highest at 30 C. 3.2. Behavior complexity based on Shannon entropy Shannon entropy based on 12 behaviors was significantly dependent on water temperature, showing the lowest value (0.29) at 15 C and the highest value at 30 C (Fig. 3a). The relative frequency of Shannon entropy was the highest between 0 and 0.2 regardless of the water temperature (Fig. 3b). However, the average frequency tended to be lower at higher temperatures (15 C: 0.71, 20 C: 0.31, 25 C: 0.18, and 30 C: 0.13). Higher entropy values and a wider entropy range (e.g., 15 C: 0.0–2.3 and 30 C: 0.0–2.8) were found at higher temperatures.
3. Results 3.3. Differences in duration of behaviors 3.1. Differences in behavior at different temperatures Test organisms spent much more time exhibiting motionless behaviors, such as clinging (38.2%), not-moving (8.3%), and floating (4.8%), at the lowest water temperature (15 C) than at higher ones (Fig. 2). Moreover, digging behavior (0.1%) was observed only at 15 C, even though the proportion was very low, whereas folding and swimming behaviors were not observed. More time was spent on feeding (23.2%) and bottom-crawling (21.6%) at 25 C than at
The frequency of short behavior duration (e.g., 1 or 2 min) was high at all water temperatures (Fig. 4). Coefficients of determination (R2) from MMF were generally higher in motion behavior (0.870–0.995) and clinging-NM (0.893–0.921) regardless of the water temperatures. The coefficients of not-moving behaviors were the lowest at all experimental temperatures (0.259–0.898) except 25 C. Duration of the same behavior became shorter, especially moving behaviors (side-crawling, bottom-crawling and
Fig. 2. Differences in behavior of golden apple snails acclimated to different water temperatures (15 C, 20 C, 25 C, and 30 C). Letters indicate significant differences among the different temperatures based on Dunn’s multiple comparison test (P < 0.05). Abbreviations of behavior categories are given in Table 1.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
5
Fig. 3. (a) Differences in Shannon entropy of snails acclimated to different water temperatures. Letters indicate significant differences among different temperatures based on Dunn's multiple comparison test (P < 0.05). (b) The relative frequency of Shannon entropy of snails acclimated to different water temperatures.
feeding behaviors), when experimental temperature was increased. This indicated that, at lower temperatures (e.g., 15 C), once a specific behavior began, it was maintained for a while before it changed to another behavior. Transition probabilities modeled with Markov chains according to the increase in behavior observation time, generally became stable within 24 h of observations (Fig. 5, Appendix A), although there was some variation in transition probability values. The transition probabilities to repeat the previous behaviors were stabilized earlier at 25 C than at other temperatures. For example, the probabilities of repeating side-crawling, bottom-crawling, feeding, clinging, and clinging-NM behaviors became static within 14 h. At the lowest temperature (15 C), despite increasing observation times, transition probabilities to repeat their previous behavior usually increased over time and stabilized at high values after 21 h. However, at the highest water temperature (30 C), transition probabilities of most behaviors were stabilized at relatively low values, except for clinging and clinging-NM behaviors, regardless of the probabilities of repeating a previous behavior or changing the type of behavior. This indicates that at high water temperatures, golden apple snails keep changing their behavior from one to another. Fig. 6 presents the transition probabilities of each behavior modeled with Markov chains based on data from 48 h of observation. The size of the boxes represents the relative ratio of time spent in a certain behavior; i.e., the larger the size, the more time spent. The line thickness represents the transition probability. Markov chains also reflected the differences in duration of behaviors. The probability of continuing a certain behavior was higher at the lowest water temperature (15 C), regardless of motion or motionless behavior, whereas the snails tended to change their behavior at higher water temperatures. For example, the probability of continuing feeding behavior was higher at 15 C (0.83), 20 C (0.85), and 25 C (0.70), than at 30 C (0.51). The probability of maintaining bottom-crawling decreased at higher water temperatures (15 C: 0.85, 20 C: 0.61, 25 C: 0.46, and 30 C: 0.30). However, the probabilities of continuing clinging (i.e., 15 C: 0.99, 20 C: 0.96, 25 C: 0.96, and 30 C: 0.80) or clinging-NM (i.e., 15 C: 0.86, 20 C: 0.82, 25 C: 0.79, and 30 C: 0.82) were high at all water temperatures, even though the probability of repeating clinging behavior slightly decreased with increasing temperature. At 15 C, the probability of showing motionless behavior (e.g., clinging) after ventilating was higher than that at
other temperatures, whereas at 30 C, the probability of showing motion behavior (e.g., side or bottom-crawling) was higher. 3.4. Evaluation of influential behaviors for predicting behavior complexity Shannon entropy was predicted based on 12 behaviors using RF modeling and the prediction power was higher at 25 C (R2: 0.45 (t + 1), 0.49 (t + 5) and 0.42 (t + 10)) than at other temperatures (15 C: 0.29 (t + 1), 0.33 (t + 5) and 0.25 (t + 10), 20 C: 0.24 (t + 1), 0.33 (t + 5) and 0.20 (t + 10) and 30 C: 0.32 (t + 1), 0.39 (t + 5) and 0.26 (t + 10)), regardless of the time (t + 1, t + 5 and t + 10) considered in the dependent variable. The important behaviors determining Shannon entropy were different at different water temperatures. However, the relative importance of behaviors was similar at the three different periods (Fig. 7). At the lowest water temperature (15 C), motion behaviors such as side-crawling (0.21) and bottom-crawling (0.18) were important to determine Shannon entropy. Clinging behavior was the most significant behavior at 20 C (0.22) and 25 C (0.27), although the importance of clinging-NM increased at 25 C (0.19). At 30 C, clinging-NM (0.24) was the most important behavior determining Shannon entropy. 4. Discussion In optimal temperature conditions (e.g., 25 C), golden apple snails regularly breathe using their siphons (Seuffert and Martín, 2010). In the present study, ventilation frequency and periodicity increased at high water temperatures, causing more changes in behavior. This was reflected on the duration of each behavior (regression analysis), the behavior transition (Markov chain analysis), and the behavior complexity (Shannon entropy). With increasing water temperatures, the duration of a certain behavior decreased, especially moving behaviors (Fig. 4) and a short (1 or 2 min) duration of a behavior was less frequent at low water temperatures. Furthermore, longest time spent in motion behavior such as crawling and feeding was observed at the lowest temperature. This indicates that, regardless of having a motion or motionless behavior, golden apple snails in the lowest water temperature treatment maintained their previous behavior (Figs. 4 and 5), although moving velocity decreased (Heiler et al., 2008).
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
Fig. 4. The relationship between behavior duration (log-transformed) and occurrence frequency (log-transformed) for the main behavior categories.
6
Transition probabilities modeled with Markov chains gave insight into the sequential predictability of golden apple snail behaviors. Results confirmed that a higher behavior transition occurred at the high water temperature (30 C) and that the transition into a motion behavior was more diverse. For example, after showing a ventilating behavior, golden apple snails at 15 C had a high probability of attaching to the side of the aquarium (clinging or clinging-NM), while snails at 30 C showed more diverse behaviors, such as bottom-crawling, side-crawling, or feeding. When organisms are exposed to harsh conditions, they try to reduce physiological and metabolic stress (Guderley, 2004). Golden apple snails displayed a digging behavior at low temperatures as a response to thermal stress, and the frequency and periodicity of the ventilating behavior were low due to motionless behavior (Fig. 2, Damborenea, 1996; Seuffert and Martín, 2009; Seuffert et al., 2010). These results corroborate earlier studies on the thermal response of golden apple snails, which have seen that they are highly sensitive to low temperatures (Seuffert and Martín, 2009; Seuffert et al., 2010). In this study, the ventilating behavior was very low at 15 C (Fig. 2). In a preliminary study, this behavior was not observed at 10 C. In addition, at a temperature between 9 C and 13 C, some golden apple snails became immobile showing a slightly protruded operculum (Stevens et al., 2002). This created water circulation and burying of the snails by siltation, without energy consumption, suggesting that this behavior may be used to minimize metabolic costs (Seuffert and Martín, 2009). These behavioral responses of snails to low temperatures allow them overwinter at open fields in Korea and expand their geographical distribution range. On the other hand, at high water temperatures, digging behavior was not observed in the present study, whereas ventilating and folding behaviors increased. At warmer water temperatures, dissolved oxygen becomes lower causing an increase in metabolic rates. As a response, golden apple snails under harsh environments minimize physiological stress through behavioral thermoregulation (increasing their ventilation frequency) (Seuffert and Martín, 2009). Similarly, snails increase their ventilation frequency when water temperature increases, leading to an increase in the oxygen-carrying capacity of the hemolymph and a higher hemocyanin level (Booth et al., 1982; Sanders and Childress, 1990). Shannon entropy quantifies the degree of repetitiveness of a certain behavior (or communicative process) according to disturbance or stress conditions (Wark et al., 2011). Humpback whales alter their feeding call patterns (increasing feeding call sequences) to be able to accurately transmit messages in the presence of vessel noise, resulting in low Shannon entropy (Doyle et al., 2008). In the current study, increases in behavior complexity (i.e., high Shannon entropy) and duration of behavior were observed at the high water temperature (30 C). At low water temperatures (15 C and 20 C), Shannon entropy was low, indicating that once a certain behavior began, it remained and did not change. However, Shannon entropy provides only the degree of behavior complexity (Wark et al., 2011), i.e., the behavior that resulted in higher or lower Shannon entropy is not known. In addition, the level of behavior turnover cannot be distinguished, and thus, similar results were obtained for behavioral sequences of high behavior turnover and low behavior turnover. Thus, the Random forest model was used to compare and analyze information on the most important behaviors determining Shannon entropy. Despite the fact that the highest frequency of clinging behavior was observed at 15 C, the importance of this behavior in determining Shannon entropy was lower than at other temperatures, while the influence of active behaviors such as bottom-crawling and side-crawling was high. This suggests that, once clinging behavior occurred, the snails
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
7
Fig. 5. Changes in behavioral transition modeled with Markov chains, according to increasing observation time. X-axis represents the observation hours (e.g., 1: Markov chain computation using data from the first hour of observation; 2: Markov chain computation using data from the first two hours of observation; up to 48: Markov chain computation using data from 48 h of observation). Y-axis represents the behavioral sequence (e.g., CB_VE means the probability of having behavior type CB at time t and VE at time t + 1). Behavioral transitions higher than 0.3 are represented in the matrix plot. Additional information on behavioral transitions is shown in the electronic supplement data.
continued showing this behavior and had a low transition probability (Figs. 3 and 4). On the other hand, as the water temperature increased, the importance of clinging-NM increased. Markov chain analysis has shown that the clinging-NM behavior is connected to most behaviors (Fig. 4), indicating that, at high water temperatures, after a clinging-NM behavior any other type of behavior can occur. When golden apple snails from tropical areas were introduced in Korea, they were not able to survive in the wild due to low
winter temperatures. However, 30 years after their first introduction, golden apple snails have adapted to the local environment (Bae et al., 2012). They are now able to overwinter in the open fields of southern part of Korea, by altering their behavioral response to low temperature. Currently, golden apple snails are commonly used for weed control in (environmentally friendly) paddy fields as substitutes for herbicides. However, golden apple snails feed vigorously on aquatic plants, not only weeds but also young rice, and a dense population of snails can
Fig. 6. Behavioral transitions based on Markov chain modeling. The size of the boxes represents the relative ratio of time spent in a certain behavior and the line thickness represents the transition probability. CL: clinging to the side of the aquarium; CBNot: not moving with a closed operculum on the bottom of the aquarium; FC: floating, with a closed operculum; CL-NM: clinging to the side of the aquarium, with spreading out antennae; CB: crawling at the bottom of the aquarium; CS: crawling at the side of the aquarium; DI: digging in the sand; FF: folding its foot to make a form like a funnel; FA: falling from the side of the aquarium; VE: breathing by lung ventilation; and FE: feeding on lettuce.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
8
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
Fig. 7. Importance of behavior variables in predicting Shannon entropy based on Random forest modelling. The values of variable importance were rescaled as the proportion of each behavior (resulting in a value of 1 when all variables were summed). Shannon entropies for t + 1, t + 5 and t + 10 were predicted from the current state of behaviors (t). Abbreviations of behavior categories are given in Table 1.
rapidly consume a large amount of plants (Lach et al., 2000; Bae et al., 2012). Bae et al. (2012) predicted that, under the influence of global warming, the geographical distribution range of these snails will increase, affecting rice production and causing adverse effects on aquatic ecosystems. However, it is difficult to prevent immediately the use of golden apple snails in environmentally friendly farming in Korea because the damage caused by these snails is not severe enough until now and the rate of weed eradication is quite high (i.e., ca. 98.6%) (Moon et al., 1997). However, it has high possibility to increase its damage in the future because of global warming as well as biological adaptation. Therefore, adequate management policies should be developed to avoid the contamination of open water systems with snails from paddy fields as well as the spread of this species. In addition, the impacts of invasive species should be fully evaluated to prevent ecosystem disturbance by invasive species when they are intended to be introduced to new area.
5. Summary and conclusions The behavioral response of golden apple snails to different acclimated water temperatures was examined based on short time interval data with various computation methods. Shannon entropy based on behaviors increased as a function of temperatures, behavior transition probability became higher at high temperature in Markov chains, and the Random forest model represented that crawling at the bottom or at side of the aquarium is more influential to determine Shannon entropy at low temperature, whereas at high temperature motionless or semi-motionless behavior showed more substantial influence at the high temperature. Golden apple snails displayed a digging behavior at low temperatures. These results revealed that golden apple snails controlled their behavior to reduce thermal stress under harsh conditions. Through this behavioral strategy, they are able to overwinter in the open fields with low temperature in Korea. Therefore, they have high potential to disperse in open water system and to be an environmental pest in Korea. Therefore, adequate management policies should be developed to minimize their ecological impacts in agriculture as well as in open aquatic ecosystems. Finally, computation methods such as Shannon entropy, Markov chain analysis, and Random forest modeling are efficient to quantify the effects of temperature changes on the behavior of golden apple snails.
Acknowledgements This work was supported by the Cooperative Research Program for Agricultural Science & Technology Development, RDA, Republic of Korea (No. PJ007420052011) and by the National Research Foundation of Korea (NRF) grant funded by the Korea government (NRF-2013R1A1A2009494). We thank two anonymous reviewers who give us constructive comments and help to improve the contents of this paper. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020. References Bae, M.J., Park, Y.S., 2014. Biological early warning system based on the responses of aquatic organisms to disturbances: a review. Sci. Total Environ. 466, 635–649. Bae, M.J., Kwon, Y.S., Park, Y.S., 2012. Effects of Global warming on the distribution of overwintering Pomacea canaliculata (Gastropoda: Ampullariidae) in Korea. Korean J. Limnol. 45, 453–458. Baker, G.H., 1998. The golden apple snail, Pomacea canaliculata (Lamarck) (Mollusca: Ampullariidae), a potential invader of freshwater habitats in Australia. In: Zalucki, M.P., Drew, R.A.I., White, G.G. (Eds.), Pest Management—Future Challenges. Proceeding of Sixth Australasian Applied Entomological Research Conference. University of Queensland, Brisbane, Australia, pp. 21–26. Baker, W.L., 1989. A review of models of landscape change. Landscape Ecol. 2, 111–133. Bang, S.W., Cho, M.K., 2008. Ecological risk of alien apple snails used in environmentally-friendly agriculture and the urgent need for its risk management in Korea. Korean J. Environ. Biol. 26, 129–137. Booth, C.E., McMahon, B.R., Pinder, A.W., 1982. Oxygen uptake and the potentiating effects of increased hemolymph lactate on oxygen transport during exercise in the blue crab, Callinectes sapidus. J. Comp. Physiol. 148, 111–121. Breiman, L., 2001. Random forests. Mach. Learn. 45, 5–32. Carlsson, N.O., Brönmark, C., Hansson, L.A., 2004. Invading herbivory: the golden apple snail alters ecosystem functioning in Asian wetlands. Ecology 85, 1575–1580. Clark, J.S., 2007. Models for Ecological Data. Princeton University Press, Princeton, New Jersey. Chon, T.S., Park, Y.S., Park, K.Y., Choi, S.Y., Kyong, T.K., Cho, E.C., 2004. Implementation of computational methods to pattern recognition of movement behavior of Blattella germanica (Blattaria Blattellidae) treated with Ca2+ signal inducing chemicals. Appl. Entomol. Zool. 39, 79–96. Cowie, R.H., Hayes, K.A., Thiengo, S.C., 2006. What are apple snails? Confused taxonomy and some preliminary resolution. In: Joshi, R.C., Sebastian, L.S. (Eds.), Global Advances in Ecology and Management of Golden Apple Snails. Philippine Rice Research Institute, Nueva Ecija, Philippines, pp. 3–24.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx Cutler, D.R., Edwards Jr., T.C., Beard, Cutler, K.H., Hess, A., Gibson, K.T., Lawler, J., 2007. Random forests for classification in ecology. Ecology 88, 2783–2792. Da Silva, M.L., Piqueira, J.R.C., Vielliard, J.M., 2000. Using Shannon entropy on measuring the individual variability in the rufous-bellied thrush Turdus rufiventris vocal communication. J. Theor. Biol. 207, 57–64. Damborenea, M.C., 1996. Patrones de distribución y abundancia de Temnocephala iheringi (Platyhelminthes, Temnocephalidae) en una población de Pomacea canaliculata (Mollusca Ampullariidae). Gayana Zool. 60, 1–12. Doyle, L.R., McCowan, B., Hanser, S.F., Chyba, C., Bucci, T., Blue, J.E., 2008. Applicability of information theory to the quantification of responses to anthropogenic noise by southeast Alaskan humpback whales. Entropy 10, 33–46. EFSA Panel on Plant Health (PLH), 2012. Scientific Opinion on the evaluation of the pest risk analysis on Pomacea insularum, the island apple snail, prepared by the Spanish Ministry of Environment and Rural and Marine Affairs. EFSA J. 10, 2552. doi:http://dx.doi.org/10.2903/j.efsa.2013.2552. Estebenet, A.L., Cazzaniga, N.J., 1992. Growth and demography of Pomacea canaliculata (Gastropoda: Ampullariidae) under laboratory conditions. Malacol. Rev. 25, 1–12. Estebenet, A.L., Martín, P.R., 2002. Pomacea canaliculata (Gastropoda: Ampullariidae): life-history traits and their plasticity. Biocell 26, 83–89. Estebenet, A.L., Martín, P.R., Silvana, B., 2006. Conchological variation in Pomacea canaliculata and other South American Ampullariidae (Caenogastropoda Architaenioglossa). Biocell 30, 329–335. Fujiwara, M., Caswell, H., 2002. Estimating population projection matrices from multi-stage mark-recapture data. Ecology 83, 3257–3265. Fukuda, S., Kang, I.J., Moroishi, J., Nakamura, A., 2010. The application of entropy for detecting behavioral responses in Japanese medaka (Oryzias latipes) exposed to different toxicants. Environ. Toxicol. 25, 446–455. Gherardi, F., Pieraccini, R., 2004. Using information theory to assess dynamics structure, and organization of crayfish agonistic repertoire. Behav. Process. 65, 163–178. Guan, D., Gao, W., Watari, K., Fukahori, H., 2008. Land use change of Kitakyushu based on landscape ecology and Markov model. J. Geogr. Sci. 18, 455–468. Guderley, H., 2004. Metabolic responses to low temperature in fish muscle. Biol. Rev. 79, 409–427. Halwart, M., 1994. The golden apple snail Pomacea canaliculata in Asian rice farming systems: present impact and future threat. Int. J. Pest. Manag. 40, 199–206. Hawkins, C.P., Cao, Y., Roper, B., 2009. Method of predicting reference condition biota affects the performance and interpretation of ecological indices. Freshwater Biol. 55, 1066–1085. Hayes, K.A., Joshi, R.C., Thiengo, S.C., Cowie, R.H., 2008. Out of South America: multiple origins of non-native apple snails in Asia. Divers. Distrib. 14, 701–712. Hazlett, B.A., Bossert, W.H., 1965. A statistical analysis of the aggressive communications systems of some hermit crabs. Anim. Behav. 13, 357–373. Heiler, K.C.M., von Oheimb, P.V., Ekschmitt, K., Albrecht, C., 2008. Studies on the temperature dependence of activity and on the diurnal activity rhythm of the invasive Pomacea canaliculata (Gastropoda: Ampullariidae). Mollusca 26, 73–81. Lach, L., Britton, D.K., Rundell, R.J., Cowie, R.H., 2000. Food preference and reproductive plasticity in an invasive freshwater snail. Biol. Invasions 2, 279–288. Liaw, A., Wiener, M., 2002. Classification and Regression by random forest. R News 2, 18–22. Lowe, S., Browne, M., Boudjelas, S., De Poorter, M., 2000. 100 of the World’s Worst Invasive Alien Species. The Invasive Species Specialist Group, IUCN, Auckland. Lusseau, D., 2003. Effects of tour boats on the behavior of bottlenose dolphins: using Markov chains to model anthropogenic impacts. Conserv. Biol. 17, 1785–1793. McCowan, B., Hanser, S.F., Doyle, L.R., 1999. Quantitative tools for comparing animal communication systems: information theory applied to bottlenose dolphin whistle repertoires. Anim. Behav. 57, 409–419.
9
Moon, Y.H., Oh, D.H., Kim, G.C., Choi, J.S., Na, J.S., 1997. Test of organic agricultural material on paddy field. Rep. Res. Exp. Chonbuk. ARES. 540–553. Mooney, H.A., Hobbs, R.J., 2000. Invasive Species in a Changing World. Island Press, Washington D.C, pp. C–$9. Morgan, P.H., Mercer, L.P., Flodin, N.W., 1975. General model for nutritional responses of higher order mechanisms. Proc. Natl. Acad. Sci. U. S. A. 72, 4327–4331. Naylor, R., 1996. Invasions in agriculture: assessing the cost of the golden apple snail in Asia. Ambio 25, 443–448. Prasad, A.M., Iverson, L.R., Liaw, A., 2006. Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems 9, 181–199. Rejesus, B.M., Sayabloc, A.S., Joshi, R.C., 1990. The distribution and control of the introduced golden apple snail (Pomacea sp.) in the Philippines, in: Introduction of Germplasm and Quarantine Procedures. PLANTI Proc. 4. PLANTI, Malaysia, 4, 213–224 Rural Development Admistration (RDA), 2007. Studies on Ecological Characteristics and Control Methods of Golden Apple Snail, Pomacea canalicula Lamarck. RDA, Korea. Sanders, Childress, J.J., 1990. Adaptations to the deep-sea oxygen minimum layer: oxygen binding by the hemocyanin of the bathypelagic mysid Gnathophausia ingens Dohrn. Biol. Bull. 178, 286–294. Seneta, E., 1966. Quasi-stationary distributions and time-reversion in genetics. J. Roy. Stat. Soc. B. Met. 28, 253–277. Seuffert, M.E., Burela, S., Martín, P.R., 2010. Influence of water temperature on the activity of the freshwater snail Pomacea canaliculata (Caenogastropoda: Ampullariidae) at its southernmost limit (Southern Pampas, Argentina). J. Therm. Biol. 35, 77–84. Seuffert, M.E., Martín, P.R., 2009. Influence of temperature, size and sex on aerial respiration of Pomacea canaliculata (Gastropoda: Ampullariidae) from southern Pampas, Argentina. Malacologia 51, 191–200. Seuffert, M.E., Martín, P.R., 2010. Dependence on aerial respiration and its influence on microdistribution in the invasive freshwater snail Pomacea canaliculata (Caenogastropoda, Ampullariidae). Biol. Invasions 12, 1695–1708. Seuffert, M.E., Martín, P.R., 2013. Juvenile growth and survival of the apple snail Pomacea canaliculata (Caenogastropoda: Ampullariidae) reared at different constant temperatures. SpringerPlus 2, 1–5. Shannon, C.E., Weaver, W., 1949. The Mathematical Theory of Information. University of Illinois Press, Illinois. StatSoft Inc., 2004. STATISTICA (data analysis software system), version 7. www. statsoft.com. Stevens, A.J., Welch, Z.C., Darby, P.C., Percival, H.F., 2002. Temperature effects on Florida apple snail activity: implications for snail kite foraging success and distribution. Wildlife Soc. B 30, 75–81. Takeda, N., 1999. Histological studies on the maturation of the reproductive system in the apple snail, Pomacea canaliculata. J. Anal. Biosci. 22, 425–432. Wada, T., 2004. Strategies for controlling the apple snail Pomacea canaliculata (Lamarck) (Gastropoda: Ampullariidae) in Japanese direct-sown paddy fields. JARQ-Jpn. Agr. Res. Quart. 38, 75–80. Wada, T., Matsukura, K., 2007. Seasonal changes in cold hardiness of the invasive freshwater apple snail, Pomacea canaliculata (Lamarck) (Gastropoda: Ampullariidae). Malacologia 49, 383–392. Wark, A.R., Wark, B.J., Lageson, T.J., Peichel, C.L., 2011. Novel methods for discriminating behavioral differences between stickleback individuals and populations in a laboratory shoaling assay. Behav. Ecol. Sociobiol. 65, 1147–1157. Winston, W.L., 1994. Operations Research Applications and Algorithms. Duxbury, California.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
Ecological Modelling xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel
Modeling behavior control of golden apple snails at different temperatures Mi-Jung Bae a , Tae-Soo Chon b , Young-Seuk Park a,c, * a
Department of Biology, Kyung Hee University, Dongdaemun, Seoul 130-701, Republic of Korea Department of Biological Sciences, Pusan National University, Geumjeong, Busan 609-735, Republic of Korea c Department of Life and Nanopharmaceutical Sciences, Kyung Hee University, Seoul 130-701, Republic of Korea b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 1 April 2014 Received in revised form 15 October 2014 Accepted 15 October 2014 Available online xxx
The golden apple snail (Pomacea canaliculata) is a detrimental invasive pest of rice in Asia. Temperature is a critical factor to determine their invasiveness including the behavior and distribution of snails. In this study, the behavioral responses of golden apple snails acclimated to different water temperatures (15 C, 20 C, 25 C, and 30 C) were examined based on Markov chain analysis, Shannon entropy, and Random forest modeling. Markov chains revealed that most snails maintained their previous behavior at low temperatures, while behavior transition tended to be higher at high temperature. Shannon entropy was also dependent on temperature (low at low temperature and high at high temperature), indicating that snails maintained their previous behavior for a long time at low temperature regardless having motion and motionless behavior, whereas they changed continuously their behavioral types at high temperature. The Random forest model showed that Shannon entropy at low temperature was influenced by crawling at the bottom or at side of the aquarium (motion behavior), while clinging to the side of the aquarium or clinging to the side of the aquarium with stretching out antennae were important behaviors determining Shannon entropy at high temperature. Our results showed that snails controlled their behavior to reduce their thermal stress and maintain a stable internal state in harsh environment. With this mechanism, they are able to overwinter in the open fields with low temperature in Korea, resulting in the increase of potential damage in agricultural ecosystems. Therefore, further study on the development of adequate management system is required to avoid their invasiveness to open water systems and ecological impacts. Finally, Shannon entropy, Markov chain, and Random forest model are useful computation methods to quantify the effects of temperature changes on the behavior of golden apple snails. ã 2014 Elsevier B.V. All rights reserved.
Keywords: Pomacea canaliculata Behavior control Random forest model Markov chain Water temperature Shannon entropy
1. Introduction Invasion by exotic species is one of the main causes of the global extinction crisis, causing biodiversity loss and biotic homogenization (Mooney and Hobbs, 2000). Once successfully introduced and established in a new area, invasive species are difficult to be eradicated. Therefore, it is essential to understand the mechanisms and processes of biological invasions. The golden apple snail (Pomacea canaliculata) is an exotic species in Asian countries, introduced from South America, which has become a serious agricultural pest, especially to young rice (Halwart, 1994; Wada,
* Corresponding author at: Department of Biology, Kyung Hee University, Dongdaemun, Seoul 130-701, Republic of Korea. Tel.: +82 2 961 0946. E-mail addresses: [email protected] (M.-J. Bae), [email protected] (T.-S. Chon), [email protected] (Y.-S. Park).
2004; Cowie et al., 2006; Hayes et al., 2008). The golden apple snail was considered a pest in the Philippines in 1986 (Rejesus et al., 1990), and since then there was an annual economic loss of about 1200 million US dollars due to aquatic crop damage (Naylor, 1996). In Japan, the distribution range of this species has increased and, in 1999, ca. 13,000 ha of paddy fields have been damaged (Wada, 2004). In addition to the severe effects of rice paddy, the snail can cause major environmental changes, such as the depletion of macrophytes in natural wetlands (Carlsson et al., 2004). Hence, the golden apple snail is indicated as one of the 100 worst invasive species worldwide by the International Union for the Conservation of Nature (Lowe et al., 2000; Bang and Cho, 2008). Temperature is a critical factor for successful establishment and colonization of invasive species, including the golden apple snail, in new areas (Baker, 1998; EFSA, 2012), owing to its effects on survival (Seuffert and Martín, 2013), growth rate (Estebenet and Cazzaniga, 1992), reproduction (Seuffert et al., 2010), lung
http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020 0304-3800/ ã 2014 Elsevier B.V. All rights reserved.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
2
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
ventilation frequency (Seuffert and Martín, 2010), feeding rate (Seuffert et al., 2010), and behavior (Heiler et al., 2008; Seuffert et al., 2010; Seuffert and Martín, 2010). Snails that overwinter in paddy fields are more tolerant to cold conditions than those that do not overwinter (Wada and Matsukura, 2007). They usually remain inactive or bury in muddy bottoms during winter or periods of cold weather conditions (Damborenea, 1996; Seuffert et al., 2010; RDA, 2007). In areas with constant warm temperatures (e.g., 25 C), females of golden apple snails reproduce only once in their lifetime (i.e., semelparous), whereas in areas with fluctuating temperature conditions (e.g., 7–24 C), they have multiple reproductive periods (i.e., iteroparous) (Estebenet and Cazzaniga, 1992; Estebenet and Martín, 2002; Seuffert et al., 2010). High movement activity and crawling velocity are observed in areas with high water temperatures (Heiler et al., 2008). However, water temperatures above 30 C were seen to cause a decrease in activity and an increase in feeding time in laboratory experiments (Seuffert et al., 2010). Animals’ behavior and physiological responses are influenced by environmental factors. Therefore, understanding the behavior of invasive species is the first step in developing an effective control and management method. However, it is difficult to analyze behavioral data of organisms as they exhibit nonlinear and complex behaviors (Bae and Park, 2014). Therefore, the analysis of such data requires the use of appropriate analytical and modeling methods, including statistical and computational approaches (Chon et al., 2004). The Shannon information theory can be used to quantify variation in behavior, duration of behavior, and behavior complexity of organisms, using different sequences of behavioral types (Fig. 1). In fact, the Shannon entropy (Shannon and Weaver, 1949) has been widely applied in animal behavior analysis, e.g., in hermit crabs (Hazlett and Bossert, 1965), dolphins (McCowan et al., 1999), the rufous-bellied thrush (Da Silva et al., 2000), and the Japanese medaka (Fukuda et al., 2010). However, only a few studies focus on behavioral sequences (Gherardi and Pieraccini, 2004). A behavioral sequence can be modeled through the Markov chain, a discrete random process in which the probability of a system in the next time step depends only on the current state of the system and is independent of the previous
step (Guan et al., 2008). The Markov chain has been used to analyze landscape changes (e.g., Baker, 1989), population ecology (e.g., Seneta, 1966; Fujiwara and Caswell, 2002), and behavioral ecology (Lusseau, 2003). Since it quantifies the dependency between the present state and the state in the next time step, it can be used to compare the probability of behavioral transitions in response to different environmental conditions (Lusseau, 2003). Most studies dealing with behavior have either used short time observations (e.g., for 2 h) or only studied a certain type of behavior at a time, such as ventilation frequency or activity. To our knowledge, there are little quantitative researches on short time interval responses in behavior. Therefore, in this study, the behavioral response of golden apple snails to different water temperatures was examined based on short time interval data (i.e., every minute). Specifically, three steps of analyses were conducted: (1) the degree of behavior complexity at different water temperatures was compared based on Shannon entropy; (2) the duration of behavior as well as behavior transitions were compared through Markov chain analysis; and (3) the factors determining behavior complexity were evaluated based on Random forest model. 2. Materials and methods 2.1. Test organisms and behavioral observations Test organisms were obtained from a farm of golden apple snails (P. canaliculata) in Korea (http://www.gpwoolunge.co.kr/). A stock population was maintained in an aquarium with dechlorinated tap water (water temperature: 25 C 1 C; L16: D8) and a 3-cm layer of sand at the bottom, for at least 2 months before starting the experiments. In order to rule out irregular behavior (e.g., spawning and mating), only males were used. These were identified based on the presence of testis, observed through the translucent shell (Takeda, 1999), or by a humped operculum (Estebenet et al., 2006) and a body size (>25 mm) (Seuffert et al., 2010). The behavior of test organisms was observed in aquaria (30 cm 30 cm) filled with water to a depth of 15 cm and with a
Fig. 1. Flow chart of behavior data analyses of golden apple snails.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
3 cm sediment layer for two days. Test organisms have been acclimated to four different temperatures (15 C, 20 C, 25 C, and 30 C) for one month prior to the experiment. The amount of lettuce fed to the snails during experiment was estimated based on the differences of lettuce leaf area before and after the experiments. Lettuce was replaced once a day in aquaria set at 15 C, 20 C, and 25 C and twice a day in the aquarium set at 30 C due to higher consumption. In each experimental condition, the behaviors of 16 test organisms were observed. The following 12 behaviors were recorded at 1-min intervals by using a web cam (C905, Logitech) (Table 1): clinging on to the side of the aquarium (clinging); not moving with a closed operculum to the bottom of the aquarium (not-moving); floating with a closed operculum (floating); clinging to the side of the aquarium with stretching out antennae (clinging-NM); crawling at the bottom of aquarium (bottom-crawling); crawling at the side of aquarium (side-crawling); digging in the sand (digging); folding its foot to make a shape like a funnel (folding); falling from the side of aquarium (falling); swimming; breathing by lung ventilation (ventilating); and feeding on lettuce (feeding). They were categorized into three groups (i.e., motionless, semi-motionless and motion) based on the degree of movement and motion behaviors (Table 1). If organisms did not move, they were considered motionless, and if they did not display any motion except the movement of their antennae, they were considered semi-motionless; other behaviors were considered as motion. 2.2. Modeling procedure 2.2.1. Descriptive measures The time spent in each type of behavior was compared between the different water temperature treatments using the Kruskal– Wallis (K–W) test and the Dunn’s multiple comparison tests, with the software Statistica (ver. 7.0; StatSoft, 2004). The duration of a certain behavior was determined based on the amount of time that golden apple snails showed uninterruptedly that behavior. To evaluate the relationship between the duration of each behavior (log transformed) and its occurrence frequency (log transformed), we tested various curve fitting regressions using a software CurveExpert 1.3 (http://curveexpert.webhop.biz/), and chose Morgan-Mercer-Flodin (MMF, Morgan et al.,1975) which showed a high fitting performance. Among the 12 categories of behavior, the six dominant behaviors (i.e., clinging, clinging-NM, bottom-crawling, side-crawling, feeding, and not-moving) were considered in this analysis. 2.2.2. Markov chain The Markov chain (also called discrete-time Markov chain or Markov model) is a stochastic process in which the probability
3
associated with transitions among different behavioral states depends solely on the current state (Clark, 2007). A change in behavioral state in the system is called a “transition,” and the probabilities associated with various state changes are called “transition probabilities.” In the Markov chain, aij is the transition probability in the system from state i at time t to state j at time t + 1. If the system has a finite number of states (1, 2, . . . , n), the Markov chain can be characterized by a transition probability matrix denoted by A in Eq. (1) (Winston, 1994):
In the present study, the transition probability matrices of the 12 categories of behavior in the four water temperature treatments were compared. In order to make sure that our observation time was enough to generalize the behavioral transition probability, Markov chains of increasing observation time were computed (i.e., Markov chain computation based on observations of 1 h, 2 h, . . . , 48 h). Markov chains were calculated in MATLAB 7. The changes in transition probabilities according to the increasing observation time were shown as the form of a matrix plot, and the Markov chain based on a 2-day observation was selected to explain the transition probabilities. 2.2.3. Shannon entropy Shannon entropy (H) takes into account the occurrence probability of each behavior as shown in Eq. (2) (Gherardi and Pieraccini, 2004): H¼
n X pðiÞlog2 pðiÞ
(2)
i¼1
where p(i) is the occurrence probability of behavior i (i.e., the occurrence frequency) and N is the number of total behaviors (N = 12). In this study, Shannon entropy was calculated using a 10-min moving average of the behavior data registered every minute. The ‘entropy’ package in R (http://cran.r-project.org) was used for this analysis. 2.2.4. Random forest Random forest (RF) model is a machine learning technique based on a combination of a large set of decision trees (Breiman, 2001). RF has no assumption concerning the relationships between predictor and response variables (e.g., linear or nonlinear), and can be applied on continuous and categorical predictors (Prasad et al., 2006; Cutler et al., 2007; Hawkins et al., 2009). A randomly selected subset of predictor variables is applied to a randomly
Table 1 Behavior categories of Pomacea canaliculata. Category
Behavior
Acronym
Motionless
Clinging to the side of the aquarium Not moving with a closed operculum on the bottom of the aquarium Floating with a closed operculum
Clinging Not-moving Floating
Semi-motionless
Clinging to the side of the aquarium with stretching out antennae
Clinging-NM
Motion
Crawling at the bottom of the aquarium Crawling at the side of the aquarium Digging in the sand Folding its foot to make a shape like a funnel Falling from the side of the aquarium Swimming Breathing by lung ventilation Feeding on lettuce
Bottom-crawling Side-crawling Digging Folding Falling Ventilating Feeding
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
4
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
selected subset of samples, minimizing overfitting and producing unbiased estimates of modeling errors. To understand the effects of the previous behavior on the variations of behavior diversity (i.e., Shannon entropy), 12 behaviors at the current time step (t) were used in this study as the independent variables, and Shannon entropy at the following time steps (e.g., t + 1, t + 5 and t + 10) was used as the dependent variable. The ‘randomForest’ package (Liaw and Wiener, 2002) in R (http://cran.r-project.org) was used for this analysis, using the default setting of three training parameters (i.e., ntree, mtry, and node size) in the RF algorithm. The variable importance was calculated using the mean decrease in accuracy. For an easier interpretation, values of variable importance were rescaled as the proportion of each behavior; therefore, the sum of variable importance values from all categories was 1.
other temperatures. Clinging-NM (22.0%) and motion behaviors including crawling (side-crawling: 33.5% and bottom-crawling: 18.3%), folding (1.3%), falling (2.2%), swimming (0.1%), and ventilating (3.3%) were the highest at 30 C. 3.2. Behavior complexity based on Shannon entropy Shannon entropy based on 12 behaviors was significantly dependent on water temperature, showing the lowest value (0.29) at 15 C and the highest value at 30 C (Fig. 3a). The relative frequency of Shannon entropy was the highest between 0 and 0.2 regardless of the water temperature (Fig. 3b). However, the average frequency tended to be lower at higher temperatures (15 C: 0.71, 20 C: 0.31, 25 C: 0.18, and 30 C: 0.13). Higher entropy values and a wider entropy range (e.g., 15 C: 0.0–2.3 and 30 C: 0.0–2.8) were found at higher temperatures.
3. Results 3.3. Differences in duration of behaviors 3.1. Differences in behavior at different temperatures Test organisms spent much more time exhibiting motionless behaviors, such as clinging (38.2%), not-moving (8.3%), and floating (4.8%), at the lowest water temperature (15 C) than at higher ones (Fig. 2). Moreover, digging behavior (0.1%) was observed only at 15 C, even though the proportion was very low, whereas folding and swimming behaviors were not observed. More time was spent on feeding (23.2%) and bottom-crawling (21.6%) at 25 C than at
The frequency of short behavior duration (e.g., 1 or 2 min) was high at all water temperatures (Fig. 4). Coefficients of determination (R2) from MMF were generally higher in motion behavior (0.870–0.995) and clinging-NM (0.893–0.921) regardless of the water temperatures. The coefficients of not-moving behaviors were the lowest at all experimental temperatures (0.259–0.898) except 25 C. Duration of the same behavior became shorter, especially moving behaviors (side-crawling, bottom-crawling and
Fig. 2. Differences in behavior of golden apple snails acclimated to different water temperatures (15 C, 20 C, 25 C, and 30 C). Letters indicate significant differences among the different temperatures based on Dunn’s multiple comparison test (P < 0.05). Abbreviations of behavior categories are given in Table 1.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
5
Fig. 3. (a) Differences in Shannon entropy of snails acclimated to different water temperatures. Letters indicate significant differences among different temperatures based on Dunn's multiple comparison test (P < 0.05). (b) The relative frequency of Shannon entropy of snails acclimated to different water temperatures.
feeding behaviors), when experimental temperature was increased. This indicated that, at lower temperatures (e.g., 15 C), once a specific behavior began, it was maintained for a while before it changed to another behavior. Transition probabilities modeled with Markov chains according to the increase in behavior observation time, generally became stable within 24 h of observations (Fig. 5, Appendix A), although there was some variation in transition probability values. The transition probabilities to repeat the previous behaviors were stabilized earlier at 25 C than at other temperatures. For example, the probabilities of repeating side-crawling, bottom-crawling, feeding, clinging, and clinging-NM behaviors became static within 14 h. At the lowest temperature (15 C), despite increasing observation times, transition probabilities to repeat their previous behavior usually increased over time and stabilized at high values after 21 h. However, at the highest water temperature (30 C), transition probabilities of most behaviors were stabilized at relatively low values, except for clinging and clinging-NM behaviors, regardless of the probabilities of repeating a previous behavior or changing the type of behavior. This indicates that at high water temperatures, golden apple snails keep changing their behavior from one to another. Fig. 6 presents the transition probabilities of each behavior modeled with Markov chains based on data from 48 h of observation. The size of the boxes represents the relative ratio of time spent in a certain behavior; i.e., the larger the size, the more time spent. The line thickness represents the transition probability. Markov chains also reflected the differences in duration of behaviors. The probability of continuing a certain behavior was higher at the lowest water temperature (15 C), regardless of motion or motionless behavior, whereas the snails tended to change their behavior at higher water temperatures. For example, the probability of continuing feeding behavior was higher at 15 C (0.83), 20 C (0.85), and 25 C (0.70), than at 30 C (0.51). The probability of maintaining bottom-crawling decreased at higher water temperatures (15 C: 0.85, 20 C: 0.61, 25 C: 0.46, and 30 C: 0.30). However, the probabilities of continuing clinging (i.e., 15 C: 0.99, 20 C: 0.96, 25 C: 0.96, and 30 C: 0.80) or clinging-NM (i.e., 15 C: 0.86, 20 C: 0.82, 25 C: 0.79, and 30 C: 0.82) were high at all water temperatures, even though the probability of repeating clinging behavior slightly decreased with increasing temperature. At 15 C, the probability of showing motionless behavior (e.g., clinging) after ventilating was higher than that at
other temperatures, whereas at 30 C, the probability of showing motion behavior (e.g., side or bottom-crawling) was higher. 3.4. Evaluation of influential behaviors for predicting behavior complexity Shannon entropy was predicted based on 12 behaviors using RF modeling and the prediction power was higher at 25 C (R2: 0.45 (t + 1), 0.49 (t + 5) and 0.42 (t + 10)) than at other temperatures (15 C: 0.29 (t + 1), 0.33 (t + 5) and 0.25 (t + 10), 20 C: 0.24 (t + 1), 0.33 (t + 5) and 0.20 (t + 10) and 30 C: 0.32 (t + 1), 0.39 (t + 5) and 0.26 (t + 10)), regardless of the time (t + 1, t + 5 and t + 10) considered in the dependent variable. The important behaviors determining Shannon entropy were different at different water temperatures. However, the relative importance of behaviors was similar at the three different periods (Fig. 7). At the lowest water temperature (15 C), motion behaviors such as side-crawling (0.21) and bottom-crawling (0.18) were important to determine Shannon entropy. Clinging behavior was the most significant behavior at 20 C (0.22) and 25 C (0.27), although the importance of clinging-NM increased at 25 C (0.19). At 30 C, clinging-NM (0.24) was the most important behavior determining Shannon entropy. 4. Discussion In optimal temperature conditions (e.g., 25 C), golden apple snails regularly breathe using their siphons (Seuffert and Martín, 2010). In the present study, ventilation frequency and periodicity increased at high water temperatures, causing more changes in behavior. This was reflected on the duration of each behavior (regression analysis), the behavior transition (Markov chain analysis), and the behavior complexity (Shannon entropy). With increasing water temperatures, the duration of a certain behavior decreased, especially moving behaviors (Fig. 4) and a short (1 or 2 min) duration of a behavior was less frequent at low water temperatures. Furthermore, longest time spent in motion behavior such as crawling and feeding was observed at the lowest temperature. This indicates that, regardless of having a motion or motionless behavior, golden apple snails in the lowest water temperature treatment maintained their previous behavior (Figs. 4 and 5), although moving velocity decreased (Heiler et al., 2008).
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
Fig. 4. The relationship between behavior duration (log-transformed) and occurrence frequency (log-transformed) for the main behavior categories.
6
Transition probabilities modeled with Markov chains gave insight into the sequential predictability of golden apple snail behaviors. Results confirmed that a higher behavior transition occurred at the high water temperature (30 C) and that the transition into a motion behavior was more diverse. For example, after showing a ventilating behavior, golden apple snails at 15 C had a high probability of attaching to the side of the aquarium (clinging or clinging-NM), while snails at 30 C showed more diverse behaviors, such as bottom-crawling, side-crawling, or feeding. When organisms are exposed to harsh conditions, they try to reduce physiological and metabolic stress (Guderley, 2004). Golden apple snails displayed a digging behavior at low temperatures as a response to thermal stress, and the frequency and periodicity of the ventilating behavior were low due to motionless behavior (Fig. 2, Damborenea, 1996; Seuffert and Martín, 2009; Seuffert et al., 2010). These results corroborate earlier studies on the thermal response of golden apple snails, which have seen that they are highly sensitive to low temperatures (Seuffert and Martín, 2009; Seuffert et al., 2010). In this study, the ventilating behavior was very low at 15 C (Fig. 2). In a preliminary study, this behavior was not observed at 10 C. In addition, at a temperature between 9 C and 13 C, some golden apple snails became immobile showing a slightly protruded operculum (Stevens et al., 2002). This created water circulation and burying of the snails by siltation, without energy consumption, suggesting that this behavior may be used to minimize metabolic costs (Seuffert and Martín, 2009). These behavioral responses of snails to low temperatures allow them overwinter at open fields in Korea and expand their geographical distribution range. On the other hand, at high water temperatures, digging behavior was not observed in the present study, whereas ventilating and folding behaviors increased. At warmer water temperatures, dissolved oxygen becomes lower causing an increase in metabolic rates. As a response, golden apple snails under harsh environments minimize physiological stress through behavioral thermoregulation (increasing their ventilation frequency) (Seuffert and Martín, 2009). Similarly, snails increase their ventilation frequency when water temperature increases, leading to an increase in the oxygen-carrying capacity of the hemolymph and a higher hemocyanin level (Booth et al., 1982; Sanders and Childress, 1990). Shannon entropy quantifies the degree of repetitiveness of a certain behavior (or communicative process) according to disturbance or stress conditions (Wark et al., 2011). Humpback whales alter their feeding call patterns (increasing feeding call sequences) to be able to accurately transmit messages in the presence of vessel noise, resulting in low Shannon entropy (Doyle et al., 2008). In the current study, increases in behavior complexity (i.e., high Shannon entropy) and duration of behavior were observed at the high water temperature (30 C). At low water temperatures (15 C and 20 C), Shannon entropy was low, indicating that once a certain behavior began, it remained and did not change. However, Shannon entropy provides only the degree of behavior complexity (Wark et al., 2011), i.e., the behavior that resulted in higher or lower Shannon entropy is not known. In addition, the level of behavior turnover cannot be distinguished, and thus, similar results were obtained for behavioral sequences of high behavior turnover and low behavior turnover. Thus, the Random forest model was used to compare and analyze information on the most important behaviors determining Shannon entropy. Despite the fact that the highest frequency of clinging behavior was observed at 15 C, the importance of this behavior in determining Shannon entropy was lower than at other temperatures, while the influence of active behaviors such as bottom-crawling and side-crawling was high. This suggests that, once clinging behavior occurred, the snails
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
7
Fig. 5. Changes in behavioral transition modeled with Markov chains, according to increasing observation time. X-axis represents the observation hours (e.g., 1: Markov chain computation using data from the first hour of observation; 2: Markov chain computation using data from the first two hours of observation; up to 48: Markov chain computation using data from 48 h of observation). Y-axis represents the behavioral sequence (e.g., CB_VE means the probability of having behavior type CB at time t and VE at time t + 1). Behavioral transitions higher than 0.3 are represented in the matrix plot. Additional information on behavioral transitions is shown in the electronic supplement data.
continued showing this behavior and had a low transition probability (Figs. 3 and 4). On the other hand, as the water temperature increased, the importance of clinging-NM increased. Markov chain analysis has shown that the clinging-NM behavior is connected to most behaviors (Fig. 4), indicating that, at high water temperatures, after a clinging-NM behavior any other type of behavior can occur. When golden apple snails from tropical areas were introduced in Korea, they were not able to survive in the wild due to low
winter temperatures. However, 30 years after their first introduction, golden apple snails have adapted to the local environment (Bae et al., 2012). They are now able to overwinter in the open fields of southern part of Korea, by altering their behavioral response to low temperature. Currently, golden apple snails are commonly used for weed control in (environmentally friendly) paddy fields as substitutes for herbicides. However, golden apple snails feed vigorously on aquatic plants, not only weeds but also young rice, and a dense population of snails can
Fig. 6. Behavioral transitions based on Markov chain modeling. The size of the boxes represents the relative ratio of time spent in a certain behavior and the line thickness represents the transition probability. CL: clinging to the side of the aquarium; CBNot: not moving with a closed operculum on the bottom of the aquarium; FC: floating, with a closed operculum; CL-NM: clinging to the side of the aquarium, with spreading out antennae; CB: crawling at the bottom of the aquarium; CS: crawling at the side of the aquarium; DI: digging in the sand; FF: folding its foot to make a form like a funnel; FA: falling from the side of the aquarium; VE: breathing by lung ventilation; and FE: feeding on lettuce.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
8
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx
Fig. 7. Importance of behavior variables in predicting Shannon entropy based on Random forest modelling. The values of variable importance were rescaled as the proportion of each behavior (resulting in a value of 1 when all variables were summed). Shannon entropies for t + 1, t + 5 and t + 10 were predicted from the current state of behaviors (t). Abbreviations of behavior categories are given in Table 1.
rapidly consume a large amount of plants (Lach et al., 2000; Bae et al., 2012). Bae et al. (2012) predicted that, under the influence of global warming, the geographical distribution range of these snails will increase, affecting rice production and causing adverse effects on aquatic ecosystems. However, it is difficult to prevent immediately the use of golden apple snails in environmentally friendly farming in Korea because the damage caused by these snails is not severe enough until now and the rate of weed eradication is quite high (i.e., ca. 98.6%) (Moon et al., 1997). However, it has high possibility to increase its damage in the future because of global warming as well as biological adaptation. Therefore, adequate management policies should be developed to avoid the contamination of open water systems with snails from paddy fields as well as the spread of this species. In addition, the impacts of invasive species should be fully evaluated to prevent ecosystem disturbance by invasive species when they are intended to be introduced to new area.
5. Summary and conclusions The behavioral response of golden apple snails to different acclimated water temperatures was examined based on short time interval data with various computation methods. Shannon entropy based on behaviors increased as a function of temperatures, behavior transition probability became higher at high temperature in Markov chains, and the Random forest model represented that crawling at the bottom or at side of the aquarium is more influential to determine Shannon entropy at low temperature, whereas at high temperature motionless or semi-motionless behavior showed more substantial influence at the high temperature. Golden apple snails displayed a digging behavior at low temperatures. These results revealed that golden apple snails controlled their behavior to reduce thermal stress under harsh conditions. Through this behavioral strategy, they are able to overwinter in the open fields with low temperature in Korea. Therefore, they have high potential to disperse in open water system and to be an environmental pest in Korea. Therefore, adequate management policies should be developed to minimize their ecological impacts in agriculture as well as in open aquatic ecosystems. Finally, computation methods such as Shannon entropy, Markov chain analysis, and Random forest modeling are efficient to quantify the effects of temperature changes on the behavior of golden apple snails.
Acknowledgements This work was supported by the Cooperative Research Program for Agricultural Science & Technology Development, RDA, Republic of Korea (No. PJ007420052011) and by the National Research Foundation of Korea (NRF) grant funded by the Korea government (NRF-2013R1A1A2009494). We thank two anonymous reviewers who give us constructive comments and help to improve the contents of this paper. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020. References Bae, M.J., Park, Y.S., 2014. Biological early warning system based on the responses of aquatic organisms to disturbances: a review. Sci. Total Environ. 466, 635–649. Bae, M.J., Kwon, Y.S., Park, Y.S., 2012. Effects of Global warming on the distribution of overwintering Pomacea canaliculata (Gastropoda: Ampullariidae) in Korea. Korean J. Limnol. 45, 453–458. Baker, G.H., 1998. The golden apple snail, Pomacea canaliculata (Lamarck) (Mollusca: Ampullariidae), a potential invader of freshwater habitats in Australia. In: Zalucki, M.P., Drew, R.A.I., White, G.G. (Eds.), Pest Management—Future Challenges. Proceeding of Sixth Australasian Applied Entomological Research Conference. University of Queensland, Brisbane, Australia, pp. 21–26. Baker, W.L., 1989. A review of models of landscape change. Landscape Ecol. 2, 111–133. Bang, S.W., Cho, M.K., 2008. Ecological risk of alien apple snails used in environmentally-friendly agriculture and the urgent need for its risk management in Korea. Korean J. Environ. Biol. 26, 129–137. Booth, C.E., McMahon, B.R., Pinder, A.W., 1982. Oxygen uptake and the potentiating effects of increased hemolymph lactate on oxygen transport during exercise in the blue crab, Callinectes sapidus. J. Comp. Physiol. 148, 111–121. Breiman, L., 2001. Random forests. Mach. Learn. 45, 5–32. Carlsson, N.O., Brönmark, C., Hansson, L.A., 2004. Invading herbivory: the golden apple snail alters ecosystem functioning in Asian wetlands. Ecology 85, 1575–1580. Clark, J.S., 2007. Models for Ecological Data. Princeton University Press, Princeton, New Jersey. Chon, T.S., Park, Y.S., Park, K.Y., Choi, S.Y., Kyong, T.K., Cho, E.C., 2004. Implementation of computational methods to pattern recognition of movement behavior of Blattella germanica (Blattaria Blattellidae) treated with Ca2+ signal inducing chemicals. Appl. Entomol. Zool. 39, 79–96. Cowie, R.H., Hayes, K.A., Thiengo, S.C., 2006. What are apple snails? Confused taxonomy and some preliminary resolution. In: Joshi, R.C., Sebastian, L.S. (Eds.), Global Advances in Ecology and Management of Golden Apple Snails. Philippine Rice Research Institute, Nueva Ecija, Philippines, pp. 3–24.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020
G Model ECOMOD 7353 No. of Pages 9
M.-J. Bae et al. / Ecological Modelling xxx (2014) xxx–xxx Cutler, D.R., Edwards Jr., T.C., Beard, Cutler, K.H., Hess, A., Gibson, K.T., Lawler, J., 2007. Random forests for classification in ecology. Ecology 88, 2783–2792. Da Silva, M.L., Piqueira, J.R.C., Vielliard, J.M., 2000. Using Shannon entropy on measuring the individual variability in the rufous-bellied thrush Turdus rufiventris vocal communication. J. Theor. Biol. 207, 57–64. Damborenea, M.C., 1996. Patrones de distribución y abundancia de Temnocephala iheringi (Platyhelminthes, Temnocephalidae) en una población de Pomacea canaliculata (Mollusca Ampullariidae). Gayana Zool. 60, 1–12. Doyle, L.R., McCowan, B., Hanser, S.F., Chyba, C., Bucci, T., Blue, J.E., 2008. Applicability of information theory to the quantification of responses to anthropogenic noise by southeast Alaskan humpback whales. Entropy 10, 33–46. EFSA Panel on Plant Health (PLH), 2012. Scientific Opinion on the evaluation of the pest risk analysis on Pomacea insularum, the island apple snail, prepared by the Spanish Ministry of Environment and Rural and Marine Affairs. EFSA J. 10, 2552. doi:http://dx.doi.org/10.2903/j.efsa.2013.2552. Estebenet, A.L., Cazzaniga, N.J., 1992. Growth and demography of Pomacea canaliculata (Gastropoda: Ampullariidae) under laboratory conditions. Malacol. Rev. 25, 1–12. Estebenet, A.L., Martín, P.R., 2002. Pomacea canaliculata (Gastropoda: Ampullariidae): life-history traits and their plasticity. Biocell 26, 83–89. Estebenet, A.L., Martín, P.R., Silvana, B., 2006. Conchological variation in Pomacea canaliculata and other South American Ampullariidae (Caenogastropoda Architaenioglossa). Biocell 30, 329–335. Fujiwara, M., Caswell, H., 2002. Estimating population projection matrices from multi-stage mark-recapture data. Ecology 83, 3257–3265. Fukuda, S., Kang, I.J., Moroishi, J., Nakamura, A., 2010. The application of entropy for detecting behavioral responses in Japanese medaka (Oryzias latipes) exposed to different toxicants. Environ. Toxicol. 25, 446–455. Gherardi, F., Pieraccini, R., 2004. Using information theory to assess dynamics structure, and organization of crayfish agonistic repertoire. Behav. Process. 65, 163–178. Guan, D., Gao, W., Watari, K., Fukahori, H., 2008. Land use change of Kitakyushu based on landscape ecology and Markov model. J. Geogr. Sci. 18, 455–468. Guderley, H., 2004. Metabolic responses to low temperature in fish muscle. Biol. Rev. 79, 409–427. Halwart, M., 1994. The golden apple snail Pomacea canaliculata in Asian rice farming systems: present impact and future threat. Int. J. Pest. Manag. 40, 199–206. Hawkins, C.P., Cao, Y., Roper, B., 2009. Method of predicting reference condition biota affects the performance and interpretation of ecological indices. Freshwater Biol. 55, 1066–1085. Hayes, K.A., Joshi, R.C., Thiengo, S.C., Cowie, R.H., 2008. Out of South America: multiple origins of non-native apple snails in Asia. Divers. Distrib. 14, 701–712. Hazlett, B.A., Bossert, W.H., 1965. A statistical analysis of the aggressive communications systems of some hermit crabs. Anim. Behav. 13, 357–373. Heiler, K.C.M., von Oheimb, P.V., Ekschmitt, K., Albrecht, C., 2008. Studies on the temperature dependence of activity and on the diurnal activity rhythm of the invasive Pomacea canaliculata (Gastropoda: Ampullariidae). Mollusca 26, 73–81. Lach, L., Britton, D.K., Rundell, R.J., Cowie, R.H., 2000. Food preference and reproductive plasticity in an invasive freshwater snail. Biol. Invasions 2, 279–288. Liaw, A., Wiener, M., 2002. Classification and Regression by random forest. R News 2, 18–22. Lowe, S., Browne, M., Boudjelas, S., De Poorter, M., 2000. 100 of the World’s Worst Invasive Alien Species. The Invasive Species Specialist Group, IUCN, Auckland. Lusseau, D., 2003. Effects of tour boats on the behavior of bottlenose dolphins: using Markov chains to model anthropogenic impacts. Conserv. Biol. 17, 1785–1793. McCowan, B., Hanser, S.F., Doyle, L.R., 1999. Quantitative tools for comparing animal communication systems: information theory applied to bottlenose dolphin whistle repertoires. Anim. Behav. 57, 409–419.
9
Moon, Y.H., Oh, D.H., Kim, G.C., Choi, J.S., Na, J.S., 1997. Test of organic agricultural material on paddy field. Rep. Res. Exp. Chonbuk. ARES. 540–553. Mooney, H.A., Hobbs, R.J., 2000. Invasive Species in a Changing World. Island Press, Washington D.C, pp. C–$9. Morgan, P.H., Mercer, L.P., Flodin, N.W., 1975. General model for nutritional responses of higher order mechanisms. Proc. Natl. Acad. Sci. U. S. A. 72, 4327–4331. Naylor, R., 1996. Invasions in agriculture: assessing the cost of the golden apple snail in Asia. Ambio 25, 443–448. Prasad, A.M., Iverson, L.R., Liaw, A., 2006. Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems 9, 181–199. Rejesus, B.M., Sayabloc, A.S., Joshi, R.C., 1990. The distribution and control of the introduced golden apple snail (Pomacea sp.) in the Philippines, in: Introduction of Germplasm and Quarantine Procedures. PLANTI Proc. 4. PLANTI, Malaysia, 4, 213–224 Rural Development Admistration (RDA), 2007. Studies on Ecological Characteristics and Control Methods of Golden Apple Snail, Pomacea canalicula Lamarck. RDA, Korea. Sanders, Childress, J.J., 1990. Adaptations to the deep-sea oxygen minimum layer: oxygen binding by the hemocyanin of the bathypelagic mysid Gnathophausia ingens Dohrn. Biol. Bull. 178, 286–294. Seneta, E., 1966. Quasi-stationary distributions and time-reversion in genetics. J. Roy. Stat. Soc. B. Met. 28, 253–277. Seuffert, M.E., Burela, S., Martín, P.R., 2010. Influence of water temperature on the activity of the freshwater snail Pomacea canaliculata (Caenogastropoda: Ampullariidae) at its southernmost limit (Southern Pampas, Argentina). J. Therm. Biol. 35, 77–84. Seuffert, M.E., Martín, P.R., 2009. Influence of temperature, size and sex on aerial respiration of Pomacea canaliculata (Gastropoda: Ampullariidae) from southern Pampas, Argentina. Malacologia 51, 191–200. Seuffert, M.E., Martín, P.R., 2010. Dependence on aerial respiration and its influence on microdistribution in the invasive freshwater snail Pomacea canaliculata (Caenogastropoda, Ampullariidae). Biol. Invasions 12, 1695–1708. Seuffert, M.E., Martín, P.R., 2013. Juvenile growth and survival of the apple snail Pomacea canaliculata (Caenogastropoda: Ampullariidae) reared at different constant temperatures. SpringerPlus 2, 1–5. Shannon, C.E., Weaver, W., 1949. The Mathematical Theory of Information. University of Illinois Press, Illinois. StatSoft Inc., 2004. STATISTICA (data analysis software system), version 7. www. statsoft.com. Stevens, A.J., Welch, Z.C., Darby, P.C., Percival, H.F., 2002. Temperature effects on Florida apple snail activity: implications for snail kite foraging success and distribution. Wildlife Soc. B 30, 75–81. Takeda, N., 1999. Histological studies on the maturation of the reproductive system in the apple snail, Pomacea canaliculata. J. Anal. Biosci. 22, 425–432. Wada, T., 2004. Strategies for controlling the apple snail Pomacea canaliculata (Lamarck) (Gastropoda: Ampullariidae) in Japanese direct-sown paddy fields. JARQ-Jpn. Agr. Res. Quart. 38, 75–80. Wada, T., Matsukura, K., 2007. Seasonal changes in cold hardiness of the invasive freshwater apple snail, Pomacea canaliculata (Lamarck) (Gastropoda: Ampullariidae). Malacologia 49, 383–392. Wark, A.R., Wark, B.J., Lageson, T.J., Peichel, C.L., 2011. Novel methods for discriminating behavioral differences between stickleback individuals and populations in a laboratory shoaling assay. Behav. Ecol. Sociobiol. 65, 1147–1157. Winston, W.L., 1994. Operations Research Applications and Algorithms. Duxbury, California.
Please cite this article in press as: Bae, M.-J., et al., Modeling behavior control of golden apple snails at different temperatures. Ecol. Model. (2014), http://dx.doi.org/10.1016/j.ecolmodel.2014.10.020